Real-Time Implementation of Face Detection and Tracking

Abstract

Face detection is a major problem in the field of image processing, and especially the requirement for speed and reliability.

In this paper, we present algorithms of face detection and tracking in a stream of video, which work in real time.

The first stage is about face detection; it has based in detection the color of human skin. In this stage, we are going to locate the faces using tow techniques of localization: projection and labeling.

The second stage is the tracking of the detected face in the following frames. The tracking operation done using the algorithm of measurement of similarity SAD (the Sum of absolute differences).

Our contribution in this article concentrates in the ameliorations of the algorithms used. Other thing is the implementation of these algorithms on the TMS320DM6437 EVM DSP platform, and the optimization level of implementation so that the system is an autonomous system, worked in real-time.

Keywords: 

Face detection and tracking, Real time,Skin color, SAD, DSP TMS320 DM6437.